A Novel Ensemble Machine Learning and Deep Learning Techniques with Feature Extraction and Selection for Fake News Stance Detection: A Comparative Analysis

In the current digital era, fake news has spread widely and may cause confusion, disinformation, and even damage to society. Accurately identifying false news positions is essential to preserving the integrity of information sharing. Deep learning (DL) methods have shown encouraging performance in a number of natural language processing applications in recent years. In order to identify false news, this work intends to investigate the efficacy of DL models and feature selection strategies. It also provides a comparative analysis of various methods. This work begins by collecting a diverse dataset comprising real news articles and their corresponding fake news counterparts and preprocess the data and extract relevant features, such as textual content, metadata, and social media engagement metrics. Next, employ various machine learning (ML) and DL models, including Ensemble Learning Model using (ELM) with Principal Component Analysis (PCA) and Fuzzy Particle Swarm Optimization (FPSO) and named as ELM-PCA-FPSO, ELM with Modified Binary Dragonfly Optimization algorithm (MBDFO) and FPSO and finally Ensemble DL models (EDLM) with Modified Binary Dragonfly Optimization algorithm (MBDFO) and Penguin Search Optimization (PSO) to learn intricate patterns and representations from the data. Compare the effectiveness of various DL models and feature extraction and selection methods via thorough testing and assessment. To evaluate the efficacy of each strategy, the study focuses on parameters like accuracy, precision, recall, and F1 score. The study's findings provide insightful information about the benefits and drawbacks of various DL approaches and feature selection strategies for the identification of Fake News Stance (FNS).